What happens when a data leader who cut his teeth at Amazon, Starbucks, Nike, and Microsoft, steps into the driver’s seat of a century-old automotive icon? You get Jon Francis, Chief Data and Analytics Officer at General Motors, navigating one of the most complex digital transformations in the industrial world.

Francis’ journey through digitally native giants shaped his perspective at GM, where AI isn’t just a buzzword, it’s the engine behind everything from hands-free driving to predictive maintenance and EV infrastructure.

But he’s quick to point out, “AI without data is just smoke in the rearview mirror,” an illusion without substance. That’s why at GM, the focus isn’t on chasing the next AI trend. It’s on anchoring every initiative in business strategy and using AI as the enabler, not the end goal.

In this interview, Francis reveals how GM is harnessing generative AI (GenAI) to modernize legacy code, deploying intelligent agents on factory floors, and pioneering data-driven innovations that are reshaping the automotive experience.

Edited Excerpts 

Q: You’ve had leadership roles at Starbucks, Amazon, Nike, Microsoft – how has that journey shaped your perspective as CDAO at GM?

It’s been an interesting journey that’s exposed me to two very different kinds of organizations. Companies like Amazon were born in the digital era, they’re engineering-led, data-driven, and built with analytics and AI at their core. Implementing AI there is more natural because it’s already part of their DNA.

In contrast, heritage brands, whether it’s GM, Nike, or Starbucks, face a different set of challenges. One of the biggest, and perhaps most surprising, is cultural. These organizations weren’t built around data, so adopting AI and analytics requires a significant shift in mindset. There’s a lot of change management involved.

At companies over 100 years old, like GM, roles and workflows have been deeply established. Bringing in AI and automation means rethinking how work gets done, freeing people from manual tasks and allowing them to focus on areas where human judgment adds the most value.

The real journey is about showing teams what’s possible and bringing them along. It’s not just about technology, it’s about helping people see how these tools can make their work more impactful and efficient. Honestly, throughout my 30-year career, that cultural shift has probably been the most eye-opening part.

Q: How has GM’s data strategy evolved, particularly in the last few years as AI has become more central to business strategy?

It’s easy to get swept up in the buzz around AI, and the media only amplifies it. But the truth is, none of it works without a strong, clean data strategy. You need reliable, production-ready data that’s timely, governed, and labeled correctly. If you’re using modern architectures, like medallion architecture, you need well-maintained silver assets and robust data quality checks.

When I joined GM, there was no shortage of big ideas about what’s possible with AI. But the real question became: what’s our data strategy? GM had made good initial moves, data was coalesced in an on-prem Hadoop environment, but what was missing was a lot of the richness around the metadata – like how do we make it really usable, the data quality checks, and also getting it close to compute.

Three years ago, we made a focused push to move our data into the cloud. That included enhancing governance, ensuring timeliness, and integrating with our AI/ML stack – Databricks on Azure.

We also built strong health checks to track data accuracy and resolve discrepancies like unexpected size or schema changes. Integrating compute and storage within Azure, and partnering closely with Microsoft, has helped us accelerate our AI applications significantly.

Q: What are the biggest challenges you face today as a data leader, not just at GM, but more broadly across the industry?

Many of the themes I’ve touched on are consistent across the industry. AI holds tremendous potential and is central to GM’s long-term strategy. It’s a key enabler for everything we do, from electric and combustion vehicles to autonomy and internal operations.

But to fully realize AI’s promise, there are four major challenges:

1. Data readiness: Before AI can be deployed at scale, foundational work must be done to ensure data is clean, reliable, governed, and accessible.
2. Change management: Driving adoption across the organization requires managing cultural shifts. In many companies, including GM, automation can be perceived as threatening. We must bring people along and show how AI enhances their roles, not replaces them.
3. Path to production: Building models in a notebook is one thing, but getting them into production, especially in complex environments like vehicles or manufacturing systems, is another. Without last-mile deployment, AI remains theoretical.
4. Scaling beyond pilots: Many companies get stuck after a successful pilot. The real challenge is scaling that success across the enterprise. Doing so demands robust MLOps, disciplined deployment practices, and a fundamentally different approach from experimentation.

Q: How do you respond to growing business pressure to adopt AI initiatives?

There’s pressure, especially when AI is dominating headlines and boardroom conversations. But that’s the wrong starting point. The right question is: What is our business strategy, and how can AI help us achieve it?

I always use the Excel analogy. No one says, “We need a Microsoft Excel strategy.” You use Excel to execute your strategy, whether it’s financial modeling, planning, or forecasting. AI should be viewed the same way. It’s an enabler, not the end goal.

At GM, we start by identifying strategic priorities across areas like manufacturing, engineering, supply chain, or customer experience. Once those are clear, we assess where AI – whether it’s machine learning, predictive analytics, or something else – can drive meaningful impact at scale.

So to me, it’s not about leading with AI. It’s about leading with strategy and letting AI support and accelerate the outcomes we’re already working toward.

We focus first on the business problem, whether that’s improving customer experience, optimizing supply chains, or advancing autonomy, and then ask, “Can AI help us do this better?” That orientation helps us avoid the shiny object syndrome and stay grounded in real business value.

Q: Can you walk us through some of GM’s most impactful AI and analytics use cases?

It’s no surprise to anyone who’s driven a GM vehicle that one of our most visible and advanced AI applications is Super Cruise, our hands-free driving technology. It’s a shining example of how we’re leveraging AI to enable more autonomous driving experiences, while still keeping the driver engaged behind the wheel.

Beyond Super Cruise, we’re also applying AI and machine learning across a wide range of use cases, both internally and customer-facing. For instance, in our manufacturing plants, we’re using machine learning-driven device-level analytics to detect potential issues proactively – whether in battery packs, conveyor belts, or other components – before they fail.

Another compelling area is around diagnostic trouble code (DTC) data. When a warning light, like a “check engine” signal, appears in a vehicle, there’s a massive payload of data behind that alert. We’re working with petabytes of this diagnostic data coming from our connected vehicles to predict potential failures before they impact the customer.

For example, much like we predict battery issues, we’re aiming to anticipate vehicle faults in advance so we can take action, either remotely or through dealerships before a customer even notices a problem.

This predictive approach not only reduces warranty-related costs for GM but more importantly, enhances customer satisfaction and loyalty. A poor ownership experience can deter someone from buying another GM vehicle.

We’re also leveraging digital twins within manufacturing to streamline and accelerate vehicle production. These virtual models allow us to simulate, test, and optimize our manufacturing processes more efficiently, supporting faster development cycles.

These are just a few of the high-impact examples where we’re deploying AI and machine learning at scale. There are many more exciting initiatives underway, some of which I, unfortunately, can’t share publicly.

Q: What about Large Language Models?

We’re using AI not to replace what we’re already good at but to enhance and scale our existing strengths. For example, we’re applying GenAI to improve how we design, test, and analyze across the product lifecycle – from software development to the manufacturing process.

Within my organization, we’re exploring agentic frameworks and co-pilot-style capabilities to automate and accelerate the data science and analytics work we do. This enables us to drive greater efficiency and scale while freeing up our teams to focus on higher-value activities.

At the same time, we’re taking a measured and thoughtful approach to building these capabilities. Responsible development is at the core of how we view the future of AI-enabled innovation.

There are also a few enterprise AI use cases worth highlighting:

• Dealer recommendations: We’ve developed a solution that helps our dealer network make smarter inventory decisions. By analyzing localized customer demand and historical data, the system provides personalized recommendations for which vehicle configurations are most likely to sell and deliver strong profitability for each specific dealer. It’s a tangible example of how AI can directly impact business outcomes.
• EV charging network planning: We’re using GenAI and advanced analytics to guide the rollout of our EV charging infrastructure. By analyzing usage patterns, geographic data, and customer needs, we’re identifying optimal locations for charging stations to better support our customers.
• Modernizing legacy data: As we continue migrating more of our workflows to the cloud, we’re encountering a significant amount of legacy code, much of it originally built for on-premises environments.

Rather than having developers manually refactor each application, we’re leveraging GenAI capabilities to accelerate and modernize that process.

These tools allow us to scale modernization efforts more efficiently, using copilot-style development support to update legacy applications in a way that aligns with modern cloud-native standards. It’s a much more scalable and effective approach than doing it all manually.

• Digitalizing manuals: Another example is our vehicle manuals, many of which are in hard copy and date back quite a ways. We’re using GenAI and chat-based capabilities to digitize and make that information more accessible. This helps not only consumers but also dealers and suppliers who need to identify and match parts or understand repair instructions.
• Enhancing data access: We’re looking at how GenAI can enhance access to legacy data as we move it through our data maturity pipeline, from bronze to silver to gold. For instance, we’re enabling non-technical users to interact with structured data through natural language queries rather than writing code. This eliminates the need for data scientists to spend time on routine data retrieval.

Q: AI agents (or Agentic AI) are generating a lot of buzz. What’s your perspective on their potential role in the automobile industry, and where do you see early opportunities for adoption? 

I am hyper-focused on starting first with the business problem and opportunity and then thinking about what tools and capabilities can be brought to bear to solve those problems. Automotive will have some natural areas for innovation in this space – from customer service, robotics within manufacturing, driver assistance technologies like GM’s Super Cruise, and even how we imagine the in-cabin experience.

In the words of Spider-Man, “With great power comes great responsibility,” and GM will be very intentional about how we bring these capabilities to life, as we put safety first in everything we do.

Q: Looking ahead, how do you see the role of the Chief Data and Analytics Officer evolving over the next 5–10 years? What emerging data and technology trends excite you the most? 

On the technology side, I am excited to see the advancement of capabilities that make it more seamless for scientists to not only build but also productionalize capabilities that, frankly, were science fiction when I started my career.

Scientists can focus more on building capabilities at the speed of the business in a way that was not previously possible. Advancements in agentic AI are what I am most excited about now to drive even more velocity and lower the bar for more scientists to also be developers.

I believe the CDAO role will continue evolving to be viewed more as a business leader role and less as a technology back-office role. The impact data, analytics, and AI can have on the business can be transformative – it will be incumbent on CDAOs to learn to speak the language of the business and lead with humility to be ultimately viewed in this light.

Some of the best CDAOs I know are viewed in this way and have the biggest impact on the companies they work in.

Q: If you had unlimited resources, what are some of the problems in the data & analytics field that you would like to solve? 

From a macro societal perspective and not specific to automotive, there are many challenges (climate change, poverty/inequality, food insecurity, mental health, etc.) that I don’t believe we’ve fully brought the power of data and analytics to help solve. Tackling even one of these problems and making a dent in a positive way to help mankind would bring me both personal and professional joy.

Q: Do you own a General Motors car? Which is your favorite GM car of all time? 

I own a Cadillac Lyriq and a GMC HUMMER EV. The Cadillac Lyriq is my favorite GM car of all time – from the quality of the vehicle to the customer-forward experience and attention to detail – not to mention I’m doing my part towards zero emissions which makes me proud to connect on a personal level to our vision statement.